Partial simultaneous updating in Hopfield memories
Bruno Cernuschi-Frías · IEEE Transactions on Systems Man and Cybernetics · 1989
A simple generalization of the Hopfield memory is presented. The model proposed updates simultaneously groups of a fixed number of neurons that are disjoint in the sense that each neuron belongs to one and only one group. An analysis is presented of the case in which one of the groups is chosen at random with equal probability and then is updated according to a rule equivalent to the one given by J.J. Hopfield (1982). It is shown that the rule minimizes an energy function in the same way as the original Hopfield model. Sufficient conditions on the corresponding connection matrix as to ensure stability are given.>